Antecedentes de la Empresa de Banquetes C&C

Empirical studies on the finance-inequality relationship started only when the Deininger and Squire’s (1996) dataset on income inequality was made available. Even with the availability of

35

income inequality datasets, empirical evidence remains scant with developed economies dominating available studies. Empirical evidence from Africa is almost non-existent, with only two peer- reviewed papers and a working paper being the known available studies (Kai and Hamori, 2009; Batuo et al., 2010 and Asongu, 2013). One can generally group the studies into two categories based on the econometric methods used. The first group employs panel data techniques in a cross- country analysis (see for instance, Li et al., 1998; Beck et al. 2004 and 2007a; Clarke et al., 2006 and 2013; Rehman et al., 2008; Kappel, 2010). The second group of studies used country-specific time series methods (e.g. Law and Tan, 2009; Law et al., 2014)

One of that earliest studies, Li et al. (1998) examined the Kuznets hypothesis, looking at the international and intertemporal variation in inequality in 49 developed and developing countries from 1947-1994 using analysis of variance (ANOVA), least square dummy variable (LSDV) and random effect (RE). They found income inequality to be stable, while income was rising for the period under study, thus rejecting the Kuznets hypothesis. Their results further suggest that the determinants of income inequality vary only slowly within countries but are significantly different across countries.

Focusing on the finance-inequality relationship, Beck et al. (2004 and 2007a) found that income inequality declines faster in countries with a well-developed financial system. Their results further suggest that well-developed financial systems induce the incomes of the poor to grow faster than the average per capita GDP growth, which lowers income inequality.

In a similarly related cross-country study Clarke et al. (2006 and 2013) investigated the relationship between finance and income inequality in 83 countries from 1960 to 1995 and recently (in their 2013 study) expanded the countries to 91 while maintaining the same period. They employed ordinary least squares (OLS) and GMM in both analyses and in the earlier study, empirical evidence strongly supports the negative linear hypothesis with some weak support for the Greenwood and Jovanovic (1990) hypothesis. Similar support for the negative linear hypothesis were found in the recent study but no support was found for Greenwood and Jovanovic (1990) while there was some modest support for the augmented Kuznets hypothesis.

Rehman et al. (2008) analysed data for 51 countries at different stages of economic growth to understand the factors driving income inequality among these groups of countries and split the data into four different income groups to test the Kuznets hypothesis. They found government spending, the literacy rate and trade openness to be the main factors driving income inequality in low, lower- middle, middle and upper income countries. Their results showed that financial development reduces income inequality regardless of the stages of economic development. They also found support for the Kuznets inverted u-shape hypothesis. However, Kappel (2010) found that

36

government spending reduces income inequality in high income but not in low-income countries. Evidence from regression analysis showed that inequality and poverty are not only reduced through better loan markets but also through well-developed stock markets. The results also identified ethnic diversity and land distribution as key factors driving income inequality.

Recently emerging evidence suggests the existence of a threshold effect of financial development and institutional quality on income inequality. For example, Kim and Lin (2011) employed an instrumental variable threshold regression approach for a panel of developed and developing countries and found evidence of a nonlinear threshold effect of financial development. Their results indicate that financial development (banks and stock markets) will disproportionately help the poor and reduce income inequality only when a country has reached a certain threshold level of financial development. Below such threshold level, financial development will hurt the poor and worsen income distribution. Tan and Law (2012) also found evidence of a below-threshold effect. Their results suggest that financial development will reduce income inequality even at the early stage of financial development but this will only be sustainable below a certain threshold level. This plays out in three phases: a phase where income inequality reduces with financial development, a phase of no change in income inequality with financial development, and the final phase of rising income inequality with further financial development, thus translating into a u-shape. Further financial development after the second phase will increase income inequality. Recently, Law et al. (2014) employed a threshold regression approach and found that financial development will reduce income inequality only after a certain level of institutional quality. They concluded that until such institutional quality has been reached, the relationship between finance and income inequality will not exist.

We now turn to studies that focused on African countries. All the studies are cross-country in approach. Kai and Hamori (2009) is the first known peer-reviewed study in Africa that examined the effect of globalisation and financial depth on income inequality in 29 SSA countries from 1980- 2002 using fixed and random effect models. Their empirical evidence showed that globalisation worsens income inequality but this effect dampens with economic development of countries. That is, since globalisation is likely to benefit those with some level of education, the equalising effect of globalisation will be higher in countries with high standards of education. Furthermore, they found that financial depth reduces income inequality but its effect declines with globalisation. That is, increased globalisation shifts financial resources towards the rich and hence the gap between the rich and the poor widens.

Batuo et al. (2010) also investigated the effect of financial development on income inequality in 22 African countries from 1980-2004 by testing the various theoretical hypotheses. They found

37

empirical support for the negative linear hypothesis that financial development reduces income inequality. Meanwhile Asongu (2013) examined the channel through which investment affects inequality and which channels are good for the poor in 13 African countries. The overall result revealed that financial development in Africa does not help the poor. The results showed that financial depth and activity reduces income inequality, whereas financial efficiency increases income inequality, providing support for Greenwood and Jovanovic’s (1990) hypothesis. That is, large average loan sizes and deposits per capita are likely to benefit the rich and well-established firms. Gries and Meierrieks (2010) also found in a group of SSA countries that weak institutional quality undermines the effectiveness of financial development to reduce income inequality in the region.

Apart from cross-country studies, there are also single country studies that have examined this dynamic relationship between financial development and income inequality. However, none of these studies looked at African countries. The negative linear hypothesis of Galor and Zeira (1993) and Banerjee and Newman (1993) enjoy overwhelming support from single country studies regardless of the method used in the analysis (see Shahbaz and Islam, 2011; Bittencourt, 2010; Liang, 2006; Hoi and Hoi, 2012). Ang (2010) found that underdevelopment of the financial sector in India hurt the poor more than the rich. Law and Tan (2009) failed to find any statistically significant effect of financial development on income inequality in Malaysia. Instead, they found a statistically significant effect of institutional quality9 _{in reducing income inequality. The findings}

also identified real GDP per capita and inflation that were statistically significant in reducing income inequality. They concluded that maintaining low inflation and improving economic development would reduce income inequality.

The foregoing review clearly shows that available empirical evidence, although not clear-cut, mostly focused on developed and non-Africandeveloping countries. Specifically, there are only two published papers from Africa. Furthermore, apart from single country studies outside Africa that have employed autoregressive distributed lag models (ARDL) in their analysis, most cross-country studies applied the conventional method of data averaging, which is not in line with empirical modelling for heterogeneous non-stationary panel data. This study argues that assuming homogeneity of slope coefficient when in fact the slopes are different may lead to misleading inferences.

9 _{Institutional quality refers to five measures of political risk services (PRS): (i) corruption, (ii) rule of law, (iii)}

bureaucratic quality, (iv) Government repudiation of contracts, and (v) risk of expropriation; and six measures from the World Governance Indicators: a) voice and accountability, b) political stability and lack of violence, c) government effectiveness, d) regulatory quality, e) rule of law, and f) control of corruption.

38

3.3 STYLISED FACTS ABOUT AFRICA

African countries remain among the poorest countries in the world and are highly unequal, with six out of the ten most unequal countries in the world in 2010 being from Africa (AfDB, 2012: 2). Besides having the lowest average per capita income amongst other regions in the world, SSA has the highest headcount poverty ratios. As shown in Figure 3.1, the headcount poverty ratio, which was 56.75% in 1990, has dropped only by 13.85% over two decades to 42.65% in 2012. In contrast, East Asia and the Pacific and South Asian, which had headcount poverty ratio above 50% in 1990, witnessed a significant drop to 7.21% and 18.75% respectively by 2012.

Figure 3.1: Poverty headcount ratio at $1.9 a day (PPP)

Source: PovcalNet, 2014

Secondly, although Africa as a whole has witnessed robust GDP growth for over a decade and a half, living standards of Africans have not improved in line with the growth in GDP. Figure 3.2 illustrates GDP per capita growth over five year intervals across regions. Figure 3.1 connects with Figure 3.2 as regions with a rapid decline in headcount poverty also showed an improvement in living standards. SSA again shows the lowest level of GDP per capita growth, which may suggest that the economic growth experienced over the past decades was not high enough to lower poverty significantly. It could also be because of economic growth being concentrated in the formal sectors while enormous untapped productive resources in the informal sectors remain excluded, thus perpetuating income inequality in the region.

39

Figure 3.2: Relative standards of living across regions 1980-2013

Source: World Bank, 2014b

Could the widening income inequality and persistent poverty rate despite robust economic growth rate be attributed to the state of the financial systems in Africa? In the past three decades, many SSA countries have adopted several financial sector reforms that emphasise market-oriented policies. For instance, in the 1980s and 1990s many of the countries in the region adopted structural adjustment programmes that emphasised the liberalisation and opening of the financial sectors as opposed to the government-controlled eras of the past. A look at indicators of financial development in the region reveals that although the region has experienced some progress in the financial sector, the sector remains largely underdeveloped and among the least developed in the world. The level of financial exclusion also remains very high, with only 35% of the adult population having access to the banking sector and other financial institutions (Global Findex, 2014). A cursory look at the indicators of financial development vis-à-vis the Gini coefficient, a measure of income inequality, in Figures 3.3 and 3.4 seems to suggest some correlation between income inequality and financial development. One can observe that in countries where domestic private credit as a ratio of GDP is rising, the Gini coefficient tends to fall. This can easily been seen in Egypt, Ethiopia, Malawi and Mauritius. Conversely, in countries where domestic private credit declines, there are also some indications that the Gini coefficient rises. This is evident in Côte d’Ivoire, Ghana, Lesotho and Mauritania. What is not clear, though, is the extent to which the level of financial development explains the behaviour of income inequality in these countries. Moreover, it is not obvious from the simple graphs whether the relationship between financial development

40

and income inequality is linear or non-linear. These can only be established using more advanced econometric techniques. Furthermore, the study performed pairwise correlation analysis for Gini coefficient and the two measures of financial development to complement Figures 3.3 and 3.4. The results reported in Table 3.1a at the end of this chapter showed that Côte d’Ivoire, Ethiopia, Nigeria and Rwanda have negative correlation of above 50% for domestic credit/GDP and 62% for bank deposits between Gini coefficients. Meanwhile, Morocco has a positive correlation of 78% between Gini coefficient and the two measures of financial development. Mauritius has on average about 50% correlation between Gini coefficient and the two measures of financial development and Malawi has a correlation of 46% for domestic credit/GDP and 91% for bank deposits/GDP between Gini coefficients. Other countries such as Egypt, South Africa and Botswana have a negative correlation of above 56% between Gini coefficients only. The correlation analysis again does not provide any strong pattern just as Figures 3.3 and 3.4. In the next section, we turn our attention to the methods that the study uses to accomplish this.

Figure 3.3: Gini net and domestic credit to GDP

0 50 100 150 0 50 100 150 0 50 100 150 0 50 100 150 1985 1990 1995 2000 2005 1985 1990 1995 2000 2005 1985 1990 1995 2000 2005 1985 1990 1995 2000 2005

Botswana Cote d'Ivoire Egypt Ethiopia

Ghana Lesotho Malawi Mauritius

Morocco Nigeria Rwanda South Africa

Tunisia Uganda

Gini net domestic credit/GDP year

Graphs by country

Source: SWIID Version 4.1 created by Solt, 2009

Gini net is the estimate of Gini index of inequality in equalised household disposable income post tax and post transfers, and Gini market (gross) is the estimate of Gini index of inequality in

41

equalised household market (pre-tax and pre-transfer) income using Luxembourg Income study data as the standard (Solt, 2014: 2).

Figure 3.4: Gini net and bank deposits/GDP

0 50 100 0 50 100 0 50 100 0 50 100 1985 1990 1995 2000 2005 1985 1990 1995 2000 2005 1985 1990 1995 2000 2005 1985 1990 1995 2000 2005

Botswana Cote d'Ivoire Egypt Ethiopia

Ghana Lesotho Malawi Mauritius

Morocco Nigeria Rwanda South Africa

Tunisia Uganda

Gini net bank deposits/GDP

year

Graphs by country

Source: SWIID Version 4.1 created by Solt, 2009

3.4 DATA DESCRIPTION

This study used more than one source for data collection. First, financial development indicators are from the World Bank global financial development (WBGFD) database. Financial development is proxied using the domestic credit to private sector (% of GDP). This ratio is widely used in the finance-growth literature (Beck et al., 2000; Beck et al., 2004). WBGFD defines this ratio as financial resources provided to the private sector by financial corporations, such as through loans, purchases of none-quality securities, and trade credits and other accounts receivable that establish a claim for payment. For some countries, these claims include credit to public enterprises. Financial corporations include monetary authorities and deposit money banks, as well as other financial corporations where data is available such as finance and leasing companies, moneylenders, insurance corporations, pension funds and foreign exchange companies. Domestic credit therefore reflects the degree to which the private sector has access to financial intermediation. A second measure of financial development used is bank deposits that represent the total value of demand, time and savings deposits at domestic money banks as a share of GDP. Deposit money banks comprise commercial banks and other financial institutions that do not accept transferable deposits

42

but incur liabilities such as time and savings deposits. These two measures are deflated by the end- of-year consumer price indices (Beck et al., 1999: 6).

Second, income inequality data are sourced from the Standardised World Income Inequality Dataset (SWIID) created by Solt (2009). The SWIID combines information from other income inequality datasets10 _{to create a standardised income inequality dataset with greater coverage that maximises}

comparability of available income inequality data for the broadest possible sample of countries and years. The SWIID uses the Luxembourg Income Study (LIS) dataset to serve as the base for standardisation (Solt, 2009: 1).

However, SWIID dataset has some limitations as discussed by Jenkins (2015), Wittenberg (2015) and Ferreira et al. (2015). This includes the strong assumption of constant ratios of Gini coefficients across series within groups of country year observations and the use of the five-year smoothing algorithm that is likely to prevent abrupt changes11_{. Despite the limitations highlighted by the above}

authors, the SWIID has been used in empirical analysis and has been published in peer-reviewed journals by researchers such as Law et al. (2014), Solt (2009), Solt et al. (2011), Solt (2015) and Sturm and De Haan (2015). Based on the research question we are trying to answer, the SWIID is the preferred secondary source data in terms of coverage, quality and comparability. The study uses control variables such as real GDP per capita measured at current United States dollars, inflation rate, consumer prices (annual %), trade openness (% of GDP), school enrolment, primary (% gross) and value added by the manufacturing sector to GDP. These variables come from the World Bank (2014b). Clarke et al. (2006 and 2013) have used similar variables in their analysis in a group of developed and developing economies. Table 3.1b at the end of this chapter depicts summary statistics of the dependent and control variables. The mean of all the variables falls between 1 and 3 with the exception of the Squared GDP per capita and inflation. Hence, there is no strong evidence of extreme observations (outliers) in the data.

10 _{The United Nations University World Income Inequality Dataset version 2.0c, the OECD Income Distribution}

Database, the OECD Income Distribution Database, the Socio-Economic Database for Latin America and the Caribbean generated by CEDLAS and the World Bank, Eurostat, the World Bank’s PovcalNet, the UN Economic Commission for Latin America and the Caribbean, the World Top Incomes Database and national statistical offices around the world.

11 _{They also indicated that the imputation procedure introduces variability in the data that needs to be accounted for in}

any empirical analysis. Jenkins (2015: 39-40) in a regression analysis illustrates that ignoring the multiple imputation when analysing the data may not lead to larger standard errors provided that the sample is drawn from the same region. The SWIID has 46 African countries with varying country and year observations. Some countries have very few observations and for the purpose of this study, we focus on countries with sufficiently long periods. This reduces the sample to only 15 African countries with complete data from 1985 to 2007. The selection criteria for the countries used in this analysis is thus based on data availability.

43

3.5 EMPIRICAL FRAMEWORK AND ECONOMETRIC SPECIFICATION

The study employs the augmented mean group (AMG) estimator that accounts for slope heterogeneity in non-stationary macro panel time series developed by (Eberhardt and Teal, 2010) and Eberhardt (2012). The AMG estimator is feasible to analyse non-stationary panel data with heterogeneous slope even if the variables are not co-integrated.

Empirical framework of heterogeneous panel

Using the Eberhardt (2012, 62) empirical modelling: for i = 1 . . ., N and t = 1… T,

it it i it βx u y   _…(3.1) where μ_it α_1i if_t ε_it …(3.2) it t i t i i it f g e x 2    …(3.3)

where x_itand yit are observables and βi is the country specific slope on the observable regressor,u_it

contains the unobservables and _it is the error term. The unobservables in Equation 3.2 comprised of country fixed effects_1i, which capture time-invariant heterogeneity across countries as well as